Luke Zettlemoyer, Margaret Li, Sneha Kudugunta
We lifted 18 functions out of this paper's own repositories and ran 14 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| hadasah/scaling_laws | canonical | 14 of 18 |
| Function | Status | Where it lives |
|---|---|---|
| adapt_df_for_isoflop | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("ba7a1602469aeabd") |
| apply_smoothing_filter | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("c10f725822634149") |
| custom_huber_loss | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("0e3ad85778a00e72") |
| fetch_flop | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("394938c35cfc309b") |
| fit_loss_with_saturation | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("a37acb41a719d908") |
| get_color | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("8d1a5df6af7bf845") |
| get_noise_for_loss | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("c2379e45b530b218") |
| maybe_get_item | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("a6d9e09c75556988") |
| minimize_with_interp | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("c569c60d28de0481") |
| power_law_fit | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("bf2301eb6d098001") |
| precise_flops_per_token_chinchilla | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("fc9bc7439b7fd07b") |
| precise_param_count_open_lm | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("e20ba47c892c9025") |
| proportional_sliding_window_filter | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("e21e1648a30e4bc5") |
| vectorized_interp_with_seed_noise | Ran | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("8d12b64a6baa8c66") |
| fit_compute_optimal_power_laws | Not yet run | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("0c16ea3bfb805dee") |
| fit_isoflop_power_law | Not yet run | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("6b4a2fc6a22b2143") |
| interp_flop | Not yet run | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("a4b194333547551a") |
| interpolation | Not yet run | hadasah/scaling_laws/paper_analysis_and_plots.py pointer only (licence: NONE) · get_code("910a5e11cfe206a8") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Modern foundation models rely heavily on using scaling laws to guide crucial training decisions. Researchers often extrapolate the optimal architecture and hyper parameters settings from smaller training runs by describing the relationship between, loss, or task performance, and scale. All components of this process vary, from the specific equation being fit, to the training setup, to the optimization method. Each of these factors may affect the fitted law, and therefore, the conclusions of a given study. We discuss discrepancies in the conclusions that several prior works reach, on questions such as the optimal token to parameter ratio. We augment this discussion with our own analysis of the critical impact that changes in specific details may effect in a scaling study, and the resulting altered conclusions. Additionally, we survey over 50 papers that study scaling trends: while 45 of these papers quantify these trends using a power law, most under-report crucial details needed to reproduce their findings. To mitigate this, we we propose a checklist for authors to consider while contributing to scaling law research.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2502.18969")
get_code_for_paper("2502.18969")
have("2502.18969")
Connect an agent — have() is free.